As of 2026-08-06, checked directly against each vendor’s own pricing page: GitHub Copilot’s first paid tier is $10 per user per month, Cursor’s is $20 per month, and Claude Code (through a Claude Pro subscription) is $20 per month billed monthly or $17 per month billed annually. Every one of the three also offers a free tier with meaningfully limited usage, and every one of the three has moved toward usage-based credits layered on top of the flat monthly price, which means the sticker price is the floor, not the whole bill. What follows is where each of those numbers came from and what actually drives a real bill above the flat rate.
Why “flat monthly price” is no longer the whole story
All three tools now combine a subscription price with a consumable credit or usage allowance rather than charging a single flat fee for unlimited use. GitHub Copilot moved its Pro, Pro+, Business, and Enterprise plans to usage-based billing on June 1, 2026, replacing the previous premium-request-unit system with credits tied directly to token consumption, though the base subscription prices themselves did not change in that transition. Cursor and Claude Code both work on a similar structure: a subscription price that includes a pool of usage, with heavier use of the most capable underlying models consuming that pool faster. The practical consequence for a real project is that a developer doing unusually heavy agentic work, large context windows, many tool calls, long sessions, can exhaust an included allowance well before the billing period ends, at which point the next dollar spent depends on the specific overage terms of that tool. How much context an assistant is actually given is one of the more direct levers on that consumption rate, since a broader, unscoped context window is not just noisier, it is also more expensive per request.
GitHub Copilot: tiers and what changed
Fetched from GitHub’s own pricing page, 2026-08-06: Free is $0 with limited monthly completions; Pro is $10 per user per month; Pro+ is $39 per user per month; Max is $100 per user per month, aimed at high-volume agent workflows. Code completions and next-edit suggestions remain unlimited on every paid plan and do not draw down the credit pool; chat and agent features do. Business and Enterprise organizational plans exist but their pricing was not displayed on the public pricing page at the time of this check GitHub directs organizations to contact sales for those figures, and that is the honest answer here rather than a guessed number.
Cursor: tiers and where the credit model differs
Fetched from Cursor’s own pricing page, 2026-08-06: the Hobby plan is free, with limited agent requests and no credit card required. Individual Pro is $20 per month. Cursor also lists Pro+ and Ultra individual tiers with higher usage allowances, and Teams plans (Standard and Premium) at $40 per user per month; the exact monthly prices for the Pro+ and Ultra individual tiers were not rendered as plain figures on the page at the time of this check, so rather than repeat a third-party estimate as fact, the accurate statement is: check Cursor’s current pricing page directly for those two tiers before budgeting against them, since the underlying credit-pool structure means the headline number and the effective cost can diverge depending on which models a project actually uses.
Claude Code, through Claude Pro or Max: tiers and what the multiplier buys
Fetched from Claude’s own pricing page, 2026-08-06: Free is $0. Pro is $20 per month billed monthly, or $17 per month with an annual subscription, and includes Claude Code access alongside the web and desktop apps. Max starts from $100 per month and offers a usage allowance several times larger than Pro’s, with a higher tier available for teams or individuals who need substantially more capacity; the exact multiplier and price point for the largest tier is best checked on Claude’s own current pricing page, since Anthropic has changed usage limits mid-year before Claude Code’s usage limits were doubled across all paid plans on 2026-05-06, which is itself evidence that these figures move and a reader should verify the current state rather than treat any single snapshot as permanent.
The cost nobody puts on the pricing page: review time
This piece deliberately stops at price, not capability which of these tools actually fits PHP and WordPress work specifically is a separate comparison, and the cheapest tier is not automatically the right one if it is a poor fit for the task. None of the three pricing pages mention what is arguably the largest real cost of adopting any of these tools on an actual project: the time spent reviewing what they produce. Industry analysis surveying AI-assisted development in 2026 has reported senior engineers spending meaningfully more time in code review than before AI-assisted coding became routine, and a measured increase in code churn, code that gets rewritten again within two weeks of being written, as AI-generated code adoption grows. Neither figure belongs on a per-seat pricing comparison, because both vary enormously by team, codebase, and how disciplined the review process already was before the tool arrived, but treating a subscription price as the full cost of adopting one of these tools understates what actually shows up in a real project’s timeline. This is presented as a factor to budget attention for, not a number to multiply into a total, because no source in this space currently ties a specific dollar figure to review overhead with the kind of rigor this cluster requires before publishing a number.
FAQ
Is the cheapest tier ever enough for real, ongoing project work?
It depends entirely on how much of the work is chat- or agent-based versus simple code completion, since completions remain unlimited on paid Copilot tiers regardless of price, while chat and agent usage draws down a credit pool that a free or entry tier caps tightly.
Why do prices differ so much between GitHub Copilot’s $10 entry tier and Cursor’s $20 entry tier?
The tiers are not measuring the same thing: Copilot’s $10 Pro tier includes a smaller credit allowance than Cursor’s $20 Pro tier, and the two tools’ underlying model access and included usage differ, so the flat monthly number alone does not tell you which is actually cheaper for a specific workload.
Does a team plan always cost more per seat than an individual plan?
Not necessarily in isolation, but team and enterprise tiers typically add centralized billing, pooled usage, and administrative controls that individual plans lack, and Copilot’s organizational tiers require contacting sales rather than a published number, which this piece does not guess at.
Will these prices still be accurate in six months?
Treat anything in this piece as pinned to 2026-08-06. GitHub Copilot changed its entire billing model in mid-2026 and Claude Code’s usage limits changed mid-year as well, so this topic sits in this cluster’s fastest-decaying group and should be re-verified against each vendor’s current page before being relied on.
Is there a way to estimate cost before committing to a paid tier?
Starting on each tool’s free tier for a real task, then noting how quickly the included allowance is consumed, is a more reliable estimate for a specific project’s actual usage pattern than any published average, since agent-heavy workflows and simple completion-heavy workflows draw down a credit pool at very different rates.
Does a subscription seat also cover the API key security question?
No. A subscription seat and a pay-as-you-go API key are different access models with different exposure, and what to do with an API key specifically is covered on its own rather than folded into a pricing comparison.
